|Publication number||US6658556 B1|
|Application number||US 09/364,288|
|Publication date||Dec 2, 2003|
|Filing date||Jul 30, 1999|
|Priority date||Jul 30, 1999|
|Publication number||09364288, 364288, US 6658556 B1, US 6658556B1, US-B1-6658556, US6658556 B1, US6658556B1|
|Inventors||Ravi Kumar Arimilli, Leo James Clark, John Steve Dodson, Guy Lynn Guthrie, Jerry Don Lewis|
|Original Assignee||International Business Machines Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (45), Non-Patent Citations (21), Referenced by (9), Classifications (24), Legal Events (5)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The present application is related to the following copending applications, which are filed on even date herewith and incorporated herein by reference:
(1) U.S. application Ser. No. 09/364,284;
(2) U.S. application Ser. No. 09/364,283;
(3) U.S. application Ser. No. 09/364,282;
(4) U.S. application Ser. No. 09/364,287;
(5) U.S. application Ser. No. 09/364,285;
(6) U.S. application Ser. No. 09/364,281; and
(7) U.S. application Ser. No. 09/364,286.
1. Technical Field
The present invention relates in general to data processing and, in particular, to the storage subsystem of a data processing system. Still more particularly, the present invention relates to a data processing system having asymmetrically partitioned load-store hardware.
2. Description of the Related Art
In order to capitalize on the high performance processing capability of a state-of-the-art processor core, the storage subsystem of a data processing system must efficiently supply the processor core with large amounts of instructions and data. Conventional data processing systems attempt to satisfy the processor core's demand for instructions and data by implementing deep cache hierarchies and wide buses capable of operating at high frequency. Although heretofore such strategies have been somewhat effective in staying apace of the demands of the core as processing frequency has increased, such strategies, because of their limited scalability, are by themselves inadequate to meet the data and instruction consumption demands of state-of-the-art and future processor technologies operating at 1 GHz and beyond.
To address the above and other shortcomings of conventional processor and data processing system architectures, the present invention introduces a processor having a hashed and partitioned storage subsystem. The processor includes at least one execution unit, an instruction sequencing unit coupled to the execution unit, and a cache subsystem including a plurality of caches that store data utilized by the execution unit. Each cache among the plurality of caches stores only data having associated addresses within a respective one of a plurality of subsets of an address space.
In one preferred embodiment, the execution units of the processor include a number of load-store units (LSUs) that each process only instructions that access data having associated addresses within a respective one of the plurality of address subsets. The load-store units can have diverse hardware such that a maximum number of instructions that can be concurrently executed is different for different load-store units or such that some of the load-store units are restricted to executing certain classes of instructions.
The processor may further be incorporated within a data processing system having a number of interconnects and a number of sets of system memory hardware that each have affinity to a respective one of the plurality of address subsets. All objects, features, and advantages of the present invention will become apparent in the following detailed written description.
The novel features believed characteristic of the invention are set forth in the appended claims. The invention itself however, as well as a preferred mode of use, further objects and advantages thereof, will best be understood by reference to the following detailed description of an illustrative embodiment when read in conjunction with the accompanying drawings, wherein:
FIG. 1 depicts an illustrative embodiment of a multiprocessor data processing system in accordance with the present invention;
FIG. 2 illustrates a more detailed block diagram of a processor in the multiprocessor data processing system of FIG. 1;
FIG. 3A depicts a circuit that can implement an exemplary hashing algorithm on selected address bits;
FIG. 3B illustrates the bit positions of the address bits forming inputs to the exemplary hashing algorithm shown in FIG. 3A;
FIGS. 4A and 4B respectively depict more detailed block diagrams of the general purpose register file (GPRF) and floating-point register file (FPRF) of the processor of FIG. 2;
FIG. 5 is a block diagram of an exemplary embodiment of a compiler in accordance with the present invention;
FIG. 6 illustrates an exemplary embodiment of an instruction within the instruction set architecture (ISA) of the processor depicted in FIG. 2; and
FIG. 7 depicts a block diagram of an illustrative embodiment of a hash selection circuit in accordance with the present invention.
With reference now to the figures and in particular with reference to FIG. 1, there is illustrated a high level block diagram of a multiprocessor data processing system in accordance with the present invention. As depicted, data processing system 8 includes a number of processors 10 a-10 d, which each comprise a single integrated circuit including a processor core and an on-chip cache subsystem, as discussed further below. Processors 10 a-10 d are all connected to each of system interconnects 12 a-12 d, which are in turn each coupled to a respective one of system memories 16 a-16 d through an associated one of memory controllers 14 a-14 d.
According to an important aspect of the present invention, data processing system 8 implements a hashed and partitioned storage subsystem. That is, instead of the single memory controller and system memory implemented in many conventional data processing systems, the present invention partitions the system memory hardware into multiple memory controllers 14 a-14 d and multiple system memories 16 a-16 d. System memories 16 a-16 d can each contain only a respective subset of all memory addresses, such that the disjoint subsets contained in all of system memories 16 together form the system memory data set. For example, each of system memories 16 may have a storage capacity of 2 GB for a total collective storage capacity of 8 GB. The subset of memory addresses assigned to each system memory 16 is determined by a hash algorithm implemented by each of processors 10 a-10 d, as discussed further below.
Systems interconnects 12 a-12 d serve as conduits for transactions between processing units 10, transactions between processing units 10 and memory controllers 14, and transactions between-processors 10 or memory controllers 14 and other snoopers (e.g., I/O controllers) that may be coupled to system interconnects 12. By virtue of the fact that each system interconnect 12 is connected to less than all of memory controllers 14 (and in the illustrated embodiment only one), each system interconnect 12 conveys only transactions that pertain to the addresses subset(s) assigned to the attached memory controller(s) 14. Advantageously, system interconnects 12, which may each be implemented as one or more buses or as a cross-point switch, can be implemented with the same or different architectures, bandwidths, and communication protocols, as will become apparent.
The hashing and partitioning of the storage subsystem of data processing system 8 is not limited in application to memory controllers 14 and system memories 16, but preferably extends to the instruction fetch units (IFUs), load-store units (LSUs), register files, and cache subsystems of processors 10. Referring now to FIG. 2, there is illustrated a high level-block diagram of a processor 10 within data processing system 8 of FIG. 1. As shown, processor 10 includes three principal collections of circuitry: instruction sequencing unit 20, execution units 22, 24 a-24 d and 26, and data storage including register files 28 and 30 and cache subsystem 32.
In the illustrative embodiment, cache subsystem 32, which provides low latency storage for data and instructions likely to be processed by the execution units of processor 10, includes level two (L2) caches 34 a-34 d and bifurcated level one (L1) instruction and data caches 36 a-36 d and 38 a-38 d, respectively. In the illustrative embodiment, L1 instruction caches 36 may be 32 kB each, L1 data caches 38 may be 16 kB each, and L2 caches 34 may be 512 kB each, for combined cache capacities of 128 kB of L1 instruction cache, 64 kB of L1 data cache, and 2 MB of L2 cache. Of course, if desired, cache subsystem 32 may also include additional levels of on-chip or off-chip in-line or lookaside caches.
As indicated by the interconnection of L1 caches 36 and 38 to respective L2 caches 34 a-34 d and the interconnection of L2caches 34 a-34 d to respective system interconnects 12 a-12 d, each L1 cache 36, 38 and each L2 cache 34 can store only data and instructions having addresses within the subset of addresses contained in system memories 16 coupled to the associated interconnect. Thus, in the illustrated example, L1 caches 36 a and 38 a and L2 cache 34 a can only cache data and instructions residing in system memory 16 a, L1 caches 36 b and 38 b and L2 cache 34 b can only cache data and instructions residing in system memory 16 b, etc.
Instruction sequencing unit 20 contains a number of instruction fetch units (IFUs) 40 a-40 d that are each coupled to a respective one of L1 instruction cache 36 a-36 d. Thus, each IFU 40 has an affinity to a particular address subset. IFUs 40 independently fetch instructions from the associated L1 instruction caches 36 and pass fetched instructions to either branch unit 42 or dispatch unit 44, depending upon whether the instructions are branch or sequential instructions, respectively. Branch instructions are processed directly by branch unit 42, but sequential instructions are opportunistically assigned by dispatch unit 44 to one of execution units 22, 24 a-24 d and 26 as execution resources (e.g., registers and a slot in completion buffer 46) become available. Dispatch unit 44 assigns instructions to execution units 22, 24 a-24 d and 26 according to instruction type and, if a load or store instruction, the target address of the instruction. In other words, integer and floating point instructions are dispatched to integer unit (IU) 22 and floating-point unit (FPU) 26, respectively, while load and store instructions are dispatched to particular ones of LSUs 24 a-24 d after dispatch unit 44 hashes the target address specified by the instruction to determine which L1 data cache 38 contains the target data. Thus, each of LSUs 24 executes only those load and store instructions targeting addresses within the particular address subset with which the associated L1 cache has affinity.
The hash algorithm implemented by dispatch unit 44, which is programmable and can be altered dynamically during operation of data processing system 8 as discussed below, can be based on any type of address (e.g., effective address, virtual address, or real (physical) address) or any combination of address types. Referring now to FIG. 3A, there is illustrated a block diagram of exemplary hashing circuit that utilizes five low order bits, which are present in effective, virtual, and real addresses, to hash an input address into one of four address subsets A-D. As shown in FIG. 3B, the five input bits, designated bits 52-56, form the high order bits of the 12-bit page offset within both the N-bit (e.g., 64-bit) effective addresses 60 utilized by processors 10 and the 42-bit real addresses 62 utilized by cache subsystem 32, memory controllers 14, and other snoopers coupled to system interconnects 12. In addition, the five selected bits form the low order bits of the index portion of the 42-bit real address 64 utilized to select a congruence class within L2 caches 34. As depicted in FIG. 3A, the exemplary hashing algorithm performs an exclusive-OR of bits 52, 54 and 56 (e.g., with an XOR gate 52) and an exclusive-OR of bits 53 and 55 (e.g., with an XOR gate 54) and decodes the two-bit result with a decoder 56 to select one of the four address subsets.
In the illustrative embodiment, dispatch unit 44 is the only point of centralization or interaction between the different instruction and data pipelines. As a consequence, if operations such as synchronizing instructions (e.g., SYNC) must be made visible to all caches or all system interconnects, dispatch unit 44 broadcasts such operations to all LSUs 24. The synchronizing instructions are thereafter made visible on all system interconnects 12.
Referring again to FIG. 2, general purpose register file (GPRF) 28 and floating-point register file (FPRF) 30 are utilized to temporarily store integer and floating-point operands consumed by and resulting from instruction execution. Thus, IU 22 is coupled to GPRF 28, FPU 26 is coupled to FPRF 30, and GPRF 28 and FPRF 30 are each coupled to one or more (and possibly all) of LSUs 24. As shown in FIGS. 4A and 4B, which respectively illustrate more detailed views of GPRF 28 and FPRF 30, each register file contains a respective set of rename registers 70, 72 for temporarily storing result data produced by the execution of instructions and a set of architected registers 74, 76 for storing operand and result data. Result data is transferred from rename registers 70, 72 to the associated set of architected registers 74, 76 following execution of an instruction under the direction of completion unit 46 within ISU 20.
In accordance with the present invention, each of rename registers 70, 72 and architected registers 74, 76 may be partitioned between the various hashes so that only result data from instructions residing at and/or targeting addresses within the subset defined by a hash are stored in rename and architected registers associated with that hash. It is important to note that the number of registers allocated to each hash within each of register sets 70, 72, 74 and 76 can differ and the number of rename and architected registers allocated to each hash may be programmable or dynamically alterable during operation of processor 10, for example, in response to an on-chip performance monitor 60 detecting a threshold number of dispatch stalls for instructions having addresses within a particular address subset.
There are several ways in which the enhanced parallelism of the hashed and partitioned storage subsystem of the present invention can be exploited. For example, a compiler can be optimized to allocate different classes of data, for example, instructions, data, and the instruction page table entries and data page table entries utilized for address translation, to different address subsets. Alternatively, the classes of data assigned to each address subset may be data for different types of applications, for example, technical or commercial. The compiler can also distribute variables accessed by software among the various address to maximize utilization of LSUs 24.
With reference now to FIG. 5, there is depicted a block diagram of an illustrative embodiment of a compiler that implements the optimizations described above. In the illustrative embodiment, compiler 80 includes a scanner/parser 82 that, in response to receipt of an instruction set architecture (ISA) source program as an input, tokenizes the ISA source program and verifies program syntax according to a defined context-free grammar. Scanner/parser 82 outputs a syntactic structure representing the program to translator 84. Translator 84 receives the output of scanner/parser 82 and generates either an intermediate code representation or target machine code after verifying that the constructs parsed by scanner/parser 82 are legal and meaningful in context. According to the illustrative embodiment, an optimizer 86 receives an intermediate code representation produced by translator 84 and optimizes the location of variables in memory, register utilization, etc., as described above by reference to a hashing algorithm known to be implemented by dispatch unit 44. The optimized intermediate code output by optimizer 86 is then utilized by machine code generator 88 to produce a target machine code executable by a processor 10.
Alternatively, or in addition to such compiler optimizations, the hashed and partitioned subsystem of the present invention can be exploited by incorporating an awareness of the hashing of memory addresses into the instruction set architecture (ISA) of processors 10. For example, FIG. 6 illustrates an ISA instruction 90 that, in addition to conventional opcode and operand fields 92 and 94, includes optional source and destination hash fields 96 and 98. Thus, a programmer could be permitted, by supplying value(s) within hash fields 96 and 98, to explicitly direct the compiler as to which address subset source data is drawn and the address subset to which result data is stored.
The above compiler and ISA mechanisms for directing data to selected address subsets are particularly advantageous when the hardware partitions having affinity with the various address subsets are individually tailored for the type and amount of data anticipated to be within each address subset. In other words, to enhance performance each hardware partition can be implemented differently from the others. For example, the hardware of some of LSUs 24 can be devoted to execution of only integer loads and stores (i.e., be connected to only GPRF 28), while the hardware of other LSUs 24 can be capable of executing only floating-point loads and stores (i.e., be connected to only FPRF 30). In addition, certain of LSUs 24 be implemented with duplicate hardware such that multiple load and store instructions targeting addresses within the address subset associated with those LSUs 24 by the hash algorithm can be executed in parallel.
Each level of cache can also be heterogeneous. For example, caches of the same type (e.g., L1 instruction cache, L1 data cache, and L2 cache) can be designed or configured with differing sizes, associativities, coherence protocols, inclusivities, sectoring, replacement policies, and prefetch behaviors. Such diversity among caches is particularly useful if different data types are allocated to different address subsets. For example, if the compiler is optimized to assign all locks to a small address subset, the caches having affinity to that address subset can be limited to a small size to reduce access latency and therefore improve system performance on updates to shared data. The “lock” caches may also exhibit a different behavior from caches associated with other address subsets, for example, a store-through (or store-with-update) rather than a write-back protocol, to make the release of a lock visible to other processors 10 via a particular system interconnect 12 in response to execution of a store-conditional instruction.
As noted above, the implementation of diverse hardware components of the same type can also extend to system interconnects 12, and can also extend to memory controllers 14 and system memories 16. For example, a particular memory controller 14 in FIG. 1 can be implemented with duplicate memory controller hardware operating in parallel, and different memory controllers 14 can access the associated system memory 16 differently to retrieve a requested cache line of data (e.g., horizontal versus vertical slicing of memory). In addition, different system memories 16 can be implemented with differing memory technologies, for example, synchronous dynamic access memory (SDRAM) versus DRAM, differing module sizes, etc.
The hashed and partitioned storage subsystem of the present invention also preferably supports dynamic hash optimization and dynamic repair capability. In a conventional processor having only one cache at each level in a cache hierarchy and single instances of other storage subsystem circuitry, the occurrence of a double-bit ECC error in a particular cache or circuit would disable the processor. In contrast, if a double-bit ECC error (which is not correctable) is detected within a particular hardware partition of a processor 10 in accordance with the present invention, the hashing algorithm implemented by dispatch unit 44 can be altered dynamically to redistribute all addresses within the address subset associated with the defective partition to one or more of the other address subsets, thus idling the defective hardware (which may also be disabled). The hashing algorithm implemented by dispatch unit 44 can also be modified to redistribute the subsets to which memory addresses belong while retaining the full number of subsets, for example, to maximize LSU utilization, to improve address bus and/or data bus utilization or to reduce single-bit (soft) errors.
With reference now to FIG. 7, there is depicted an exemplary embodiment of a hash selection circuit that supports dynamic changes to the hashing algorithm implemented by dispatch unit 44. Hash selection circuit 100 includes a number of hashing circuits 102 (one of which may be hashing circuit 50 of FIG. 3A) that each receive certain of the bits of effective address 60 as inputs and provide a hash output designating one of the hardware partitions. As noted above, the hashing algorithms implemented by hashing circuits 102 preferably differ, such that some of hashing circuits 102 hash addresses to fewer than all of the hardware partitions and others of hashing circuits 102 provide different hashes but still distribute addresses among all hardware partitions. The hash output of each hashing circuit 102 forms an input of multiplexer 104, which selects one of the hash outputs as its output in response to select signal 106. As illustrated, select signal 106 is derived from the contents of control register 108, which may in turn be set by either or both of monitoring software and monitoring hardware (e.g., performance monitor 60).
Once a dynamic update has been made to the control register 108 of a processor 10 within data processing system 8, coherent operation of data processing system 8 requires that a similar update be performed at each of the other processors 10. These updates can be handled by sourcing special transactions from the updated processor 10 on system interconnects 12 or by execution of a software exception handler that writes a new value to each other control register 108. To minimize the performance penalty associated with a dynamic hash update, L2 caches 34 are preferably implemented such that full addresses are utilized and such that each L2 cache 34 snoops all address transactions regardless of the address subset to which the address transactions belong. With this arrangement, a dynamic change in the address hash implemented by dispatch unit 44 would require only caches in a disabled hardware partition to be flushed. However, if each L2 cache 34 only snoops address transactions for its assigned address subset, all LSUs 24 and caches within each hardware partition from which any address is reassigned would have to be flushed prior to enforcing a dynamic change in the hash.
As has been described, the present invention provides an improved processor and data processing system architecture having a hashed and partitioned storage subsystem. The present invention not only enhances performance through increased hardware parallelism, but also permits the various hardware partitions to be individually optimized for the type of data contained in each address subset. Advantageously, the address subset assigned to each hardware partition can be changed dynamically by updating the hash, thus permitting runtime optimization and dynamic repair capability. The hashed and partitioned architecture of the present invention is also highly scalable and supports future increases in processor operating frequency through the addition of more hardware partitions.
While the invention has been particularly shown and described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention. For example, although a compiler in accordance with the present invention can reside within the volatile and non-volatile storage of an operating data processing system, the compiler may alternatively be implemented as a program product for use with a data processing system. Such a program product can be delivered to a data processing system via a variety of signal-bearing media, which include, without limitation, non-rewritable storage media (e.g., CD-ROM), rewritable storage media (e.g., a floppy diskette or hard disk drive), and communication media, such as digital and analog networks. It should be understood, therefore, that such signal-bearing media, when carrying or encoding computer readable instructions that direct the functions of the present invention, represent alternative embodiments of the present invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4646237 *||Dec 5, 1983||Feb 24, 1987||Ncr Corporation||Data handling system for handling data transfers between a cache memory and a main memory|
|US4755930 *||Jun 27, 1985||Jul 5, 1988||Encore Computer Corporation||Hierarchical cache memory system and method|
|US4933940||May 12, 1989||Jun 12, 1990||Allied-Signal Inc.||Operations controller for a fault tolerant multiple node processing system|
|US5307506||Sep 16, 1992||Apr 26, 1994||Digital Equipment Corporation||High bandwidth multiple computer bus apparatus|
|US5404469 *||Feb 25, 1992||Apr 4, 1995||Industrial Technology Research Institute||Multi-threaded microprocessor architecture utilizing static interleaving|
|US5418922||Apr 30, 1992||May 23, 1995||International Business Machines Corporation||History table for set prediction for accessing a set associative cache|
|US5463739||Dec 22, 1992||Oct 31, 1995||International Business Machines Corporation||Apparatus for vetoing reallocation requests during a data transfer based on data bus latency and the number of received reallocation requests below a threshold|
|US5499350 *||Jun 6, 1995||Mar 12, 1996||Fujitsu Limited||Vector data processing system with instruction synchronization|
|US5530881 *||Jun 5, 1992||Jun 25, 1996||Hitachi, Ltd.||Vector processing apparatus for processing different instruction set architectures corresponding to mingled-type programs and separate-type programs|
|US5542066||Dec 23, 1993||Jul 30, 1996||International Business Machines Corporation||Destaging modified data blocks from cache memory|
|US5546535||Jan 13, 1995||Aug 13, 1996||Emc Corporation||Multiple controller sharing in a redundant storage array|
|US5649090||May 31, 1991||Jul 15, 1997||Bull Hn Information Systems Inc.||Fault tolerant multiprocessor computer system|
|US5694567 *||Feb 9, 1995||Dec 2, 1997||Integrated Device Technology, Inc.||Direct-mapped cache with cache locking allowing expanded contiguous memory storage by swapping one or more tag bits with one or more index bits|
|US5713001||Aug 31, 1995||Jan 27, 1998||International Business Machines Corporation||Circuit for converting address operands supplied by a program to hashed virtual address|
|US5717893||May 20, 1991||Feb 10, 1998||International Business Machines Corporation||Method for managing a cache hierarchy having a least recently used (LRU) global cache and a plurality of LRU destaging local caches containing counterpart datatype partitions|
|US5754844||Dec 14, 1995||May 19, 1998||Sun Microsystems, Inc.||Method and system for accessing chunks of data using matching of an access tab and hashing code to generate a suggested storage location|
|US5809494||Nov 16, 1995||Sep 15, 1998||Applied Language Technologies, Inc.||Method for rapidly and efficiently hashing records of large databases|
|US5854913||Jun 10, 1997||Dec 29, 1998||International Business Machines Corporation||Microprocessor with an architecture mode control capable of supporting extensions of two distinct instruction-set architectures|
|US5860101 *||Dec 17, 1997||Jan 12, 1999||International Business Machines Corporation||Scalable symmetric multiprocessor data-processing system with data allocation among private caches and segments of system memory|
|US5893163||Dec 17, 1997||Apr 6, 1999||International Business Machines Corporation||Method and system for allocating data among cache memories within a symmetric multiprocessor data-processing system|
|US5897655||Dec 10, 1996||Apr 27, 1999||International Business Machines Corporation||System and method for cache replacement within a cache set based on valid, modified or least recently used status in order of preference|
|US5903769 *||Mar 31, 1997||May 11, 1999||Sun Microsystems, Inc.||Conditional vector processing|
|US5909694||Jun 12, 1997||Jun 1, 1999||International Business Machines Corporation||Multiway associative external microprocessor cache|
|US5933860||Jul 29, 1997||Aug 3, 1999||Digital Equipment Corporation||Multiprobe instruction cache with instruction-based probe hint generation and training whereby the cache bank or way to be accessed next is predicted|
|US5940877||Jun 12, 1997||Aug 17, 1999||International Business Machines Corporation||Cache address generation with and without carry-in|
|US5943684||Apr 14, 1997||Aug 24, 1999||International Business Machines Corporation||Method and system of providing a cache-coherency protocol for maintaining cache coherency within a multiprocessor data-processing system|
|US5956751||Dec 14, 1995||Sep 21, 1999||Industrial Technology Research Institute||Computer memory management system using physical segmentation space allocation|
|US5974538||Feb 21, 1997||Oct 26, 1999||Wilmot, Ii; Richard Byron||Method and apparatus for annotating operands in a computer system with source instruction identifiers|
|US5996060 *||Sep 25, 1997||Nov 30, 1999||Technion Research And Development Foundation Ltd.||System and method for concurrent processing|
|US6014730||Dec 18, 1997||Jan 11, 2000||Nec Corporation||Dynamic adding system for memory files shared among hosts, dynamic adding method for memory files shared among hosts, and computer-readable medium recording dynamic adding program for memory files shared among hosts|
|US6038647 *||Jun 12, 1996||Mar 14, 2000||Fujitsu Limited||Cache memory device and method for providing concurrent independent multiple accesses to different subsets within the device|
|US6081880||Mar 9, 1995||Jun 27, 2000||Lsi Logic Corporation||Processor having a scalable, uni/multi-dimensional, and virtually/physically addressed operand register file|
|US6094710||Dec 17, 1997||Jul 25, 2000||International Business Machines Corporation||Method and system for increasing system memory bandwidth within a symmetric multiprocessor data-processing system|
|US6128641||Sep 12, 1997||Oct 3, 2000||Siemens Aktiengesellschaft||Data processing unit with hardware assisted context switching capability|
|US6138209||Sep 5, 1997||Oct 24, 2000||International Business Machines Corporation||Data processing system and multi-way set associative cache utilizing class predict data structure and method thereof|
|US6141707||May 28, 1998||Oct 31, 2000||Emc Corporation||Input/output request allocation by establishing master command queue among plurality of command queues to receive and store commands, determine logical volume, and forwarding command to determined logical volume|
|US6145054||Jan 21, 1998||Nov 7, 2000||Sun Microsystems, Inc.||Apparatus and method for handling multiple mergeable misses in a non-blocking cache|
|US6161166||Mar 10, 1999||Dec 12, 2000||International Business Machines Corporation||Instruction cache for multithreaded processor|
|US6175770||Dec 31, 1997||Jan 16, 2001||Dana Corporation||Electronic controller having automatic self-configuration capabilities|
|US6199179||Jun 10, 1998||Mar 6, 2001||Compaq Computer Corporation||Method and apparatus for failure recovery in a multi-processor computer system|
|US6205519||May 27, 1998||Mar 20, 2001||Hewlett Packard Company||Cache management for a multi-threaded processor|
|US6209059||Sep 25, 1997||Mar 27, 2001||Emc Corporation||Method and apparatus for the on-line reconfiguration of the logical volumes of a data storage system|
|US6240487||Feb 18, 1998||May 29, 2001||International Business Machines Corporation||Integrated cache buffers|
|US6275919||Oct 15, 1998||Aug 14, 2001||Creative Technology Ltd.||Memory storage and retrieval with multiple hashing functions|
|US6282630 *||Sep 10, 1999||Aug 28, 2001||Seiko Epson Corporation||High-performance, superscalar-based computer system with out-of-order instruction execution and concurrent results distribution|
|1||Cho, Sangyeun et al., "Decoupling Local Variable Accesses in a Wide-Issue Superscalar Processor," IEEE Proceedings of the 26<th >International Symposium on Computer Architecture, 1999, pp. 100-110.|
|2||Cho, Sangyeun et al., "Decoupling Local Variable Accesses in a Wide-Issue Superscalar Processor," IEEE Proceedings of the 26th International Symposium on Computer Architecture, 1999, pp. 100-110.|
|3||*||Graf, Rudolf F., Modern Dictionary of Electronics, 1988, Howard W. Sams & Company, Sixth Edition, p. 53.*|
|4||Gupta et al., "Demonstration of Automatic Data Partitioning Techniques for Parallelizing Compilers on Multicomputers," IEEE, pp. 179-193, Mar. 1992.|
|5||Leupers et al., "Retargetable Assembly Code Generation by Bootstrapping," IEEE, pp.: 88-93, May 1994.|
|6||Liem et al., "Address Calculation for Retargetable Compilation and Exploration of Instruction-Set Architectures," ACM, pp. 597,600, Jun. 1996.|
|7||Marwedel, "Code Generation for Core Processors" IEEE, pp. 232-237, Jun. 1997.|
|8||Neefs, Hank et al., "A Technique for High Bandwidth and Deterministic Low Latency Load/Store Accesses to Multiple Cache Banks," IEEE Sixth International Symposium on High-Performance Computer Architecture, 1999, pp. 313-324.|
|9||*||Neefs, Hank et al., "A Technique for High Bandwidth and Deterministic Low Latency Load/Store Accesses to Multiple Cache Banks," IEEE Sixth International Symposium on High-Performance Computer Architecture, 1999, pp. 313-324.*|
|10||*||Sangyeun, Cho et al., "Decoupling Local Variable Accesses in a Wide-Issue Superscalar Processor," IEEE Proceedings of the 26th International Symposium on Computer Architecture, 1999, pp. 100-110.*|
|11||U.S. patent application Ser. No. 09/364,281, Ravi K. Arimilli, et al., filed Jul. 30, 1999.|
|12||U.S. patent application Ser. No. 09/364,282, Ravi K. Arimilli, et al., filed Jul. 30, 1999.|
|13||U.S. patent application Ser. No. 09/364,283, Ravi K. Arimilli, et al., filed Jul. 30, 1999.|
|14||U.S. patent application Ser. No. 09/364,284, Ravi K. Arimilli, et al., filed Jul. 30, 1999.|
|15||U.S. patent application Ser. No. 09/364,285, Ravi K. Arimilli, et al., filed Jul. 30, 1999.|
|16||U.S. patent application Ser. No. 09/364,286, Ravi K. Arimilli, et al., filed Jul. 30, 1999.|
|17||U.S. patent application Ser. No. 09/364,287, Ravi K. Arimilli, et al., filed Jul. 30, 1999.|
|18||Yoaz, Adi et al., "Speculation Techniques for Improving Load Related Instruction Scheduling," IEEE Proceedings on the 26<th >International Symposium on Computer Architecture, 1999, pp. 42-53.|
|19||*||Yoaz, Adi et al., "Speculation Techniques for Improving Load Related Instruction Scheduling," IEEE Proceedings on the 26<th >International Symposium on Computer Architecture, 1999, pp. 42-53.*|
|20||Yoaz, Adi et al., "Speculation Techniques for Improving Load Related Instruction Scheduling," IEEE Proceedings on the 26th International Symposium on Computer Architecture, 1999, pp. 42-53.|
|21||Yoaz, Adi et al., "Speculation Techniques for Improving Load Related Instruction Scheduling," IEEE Proceedings on the 26th International Symposium on Computer Architecture, 1999, pp. 42-53.*|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7480771 *||Aug 17, 2006||Jan 20, 2009||Sun Microsystems, Inc.||Conditional synchronization mechanisms allowing multiple store operations to become visible while a flagged memory location is owned and remains unchanged|
|US7925640 *||Apr 12, 2011||Oracle America, Inc.||Dynamic multiple inheritance method dispatch data structure including an m-table size, i-table containing one or more holder addressor regions and type extension testing by frugal perfect hashing|
|US8341422||Jul 20, 2006||Dec 25, 2012||Apple Inc.||Method and apparatus for incremental code signing|
|US8364965 *||Mar 15, 2006||Jan 29, 2013||Apple Inc.||Optimized integrity verification procedures|
|US8886947||Dec 20, 2012||Nov 11, 2014||Apple Inc.||Optimized integrity verification procedures|
|US20060288223 *||Jul 20, 2006||Dec 21, 2006||Perry Kiehtreiber||Method and Apparatus for Incremental Code Signing|
|US20070043915 *||Aug 17, 2006||Feb 22, 2007||Sun Microsystems, Inc.||Conditional multistore synchronization mechanisms|
|US20070220261 *||Mar 15, 2006||Sep 20, 2007||Farrugia Augustin J||Optimized integrity verification procedures|
|US20090210379 *||Feb 14, 2008||Aug 20, 2009||Sun Microsystems, Inc.||Dynamic multiple inheritance method dispatch and type extension testing by frugal perfect hashing|
|U.S. Classification||712/225, 711/216, 712/200, 712/E09.071, 712/E09.055, 712/E09.046, 712/E09.027, 712/213|
|International Classification||G06F9/38, G06F9/30|
|Cooperative Classification||G06F9/3885, G06F9/3824, G06F9/3013, G06F9/3802, G06F9/384, G06F9/3814, G06F9/3891|
|European Classification||G06F9/38T6C, G06F9/38B, G06F9/38E1R, G06F9/38B8, G06F9/30R5D, G06F9/38D, G06F9/38T|
|Jul 30, 1999||AS||Assignment|
Owner name: INTERNATIONAL BUSINESS MACHINES CORPORATION, NEW Y
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:ARIMILLI, RAVI K.;CLARK, LEO J.;DODSON, JOHN S.;AND OTHERS;REEL/FRAME:010141/0461
Effective date: 19990728
|Jan 10, 2007||FPAY||Fee payment|
Year of fee payment: 4
|Jul 11, 2011||REMI||Maintenance fee reminder mailed|
|Dec 2, 2011||LAPS||Lapse for failure to pay maintenance fees|
|Jan 24, 2012||FP||Expired due to failure to pay maintenance fee|
Effective date: 20111202